Super-Resolution Fluorescence Microscopy

Part 1: Super-Resolution Fluorescence Microscopy

00:00:00.28 Hello. Welcome to my iBioSeminar series. My name Xiaowei Zhuang,
00:00:07.04 and I am a professor in the Department of Chemistry and Chemical Biology
00:00:13.02 and the Department of Physics at Harvard University.
00:00:17.19 I am also an Investigator at Howard Hughes Medical Institute.
00:00:22.07 In this seminar series I will describe a few fluorescence imaging and spectroscopy methods
00:00:30.09 that can help us visualize biological systems with a nanoscopic or nanometer scale resolution.
00:00:39.10 And in Part One, I will give an introduction to the recently developed super-resolution fluorescence microscopy,
00:00:49.20 which breaks the diffraction limit
00:00:52.16 with special emphasis on the stochastic optical reconstruction microscopy that my lab developed.
00:00:58.28 And in Part Two, I will give a couple of example applications of STORM that has already generated new biological insights.
00:01:09.11 And then in Part Three I will describe a single molecule spectroscopy tool
00:01:15.05 namely single molecule FRET and a couple of applications of single molecule FRET.
00:01:21.20 So as we all know, the living world encompasses an enormous range of length scales
00:01:32.18 that starts, for example, with the size of a human being of about a meter,
00:01:41.06 all the way down to the size of a protein, a few nanometers.
00:01:48.16 So of course over this enormous range of length scales we couldn't hope for a single imaging technique
00:01:56.11 that covers it all. So for that purpose there are a variety of bioimaging tools or technologies
00:02:06.15 that have been developed.
00:02:08.00 For example in the macroscopic domain we have powerful tools such as PET, MRI, and computer aided tomography,
00:02:21.08 and these tools have not only revealed important fundamental biological insights,
00:02:26.19 but have also been enormously useful for medical diagnosis.
00:02:30.27 And in the microscopic domain, we have tools such as light microscopy,
00:02:39.28 that have helped us tremendously in investigating biological systems at the cellular or even subcellular level.
00:02:49.13 And if we go to the nanoscopic domain, we have tools such as electron microscopy and spinning probe microscopy,
00:03:00.24 again, very powerful in revealing new biological insights.
00:03:05.12 And finally if we are interested in looking into the atomic arrangement of biomolecules
00:03:12.22 we have these powerful tools such as X-ray crystallography, NMR,
00:03:18.26 and also nowadays very high resolution electron microscopy.
00:03:24.10 And as I have already mentioned, light microscopy has been a workhorse for cell biological research.
00:03:34.19 And two of the advantages of light microscopy is really well re-capitulated by this movie
00:03:43.14 that I downloaded from the book of Molecular Biology of the Cell by Bruce Alberts, et al.
00:03:52.08 So what you see here are dividing embryonic cells, and the different colors, blue stains for DNA,
00:04:01.10 and the green stains for microtubules.
00:04:03.13 So because light is relatively non-invasive - it gives a pretty low perturbation to living systems.
00:04:12.17 We all know that we can go outside and enjoy the sunshine without worrying too much about damage.
00:04:19.12 Of course, when the sun is too bright as now in Woods Hole, you'll want to put on some sunscreen.
00:04:24.27 But otherwise it is pretty non-invasive approach that allows us to monitor the real time dynamics of living organisms.
00:04:35.04 Now another advantage is also well illustrated by the fact that in this movie you see two different colored objects.
00:04:44.24 And because light is so wonderfully colorful,
00:04:47.08 and because we have so many different colored fluorescence probes,
00:04:51.01 for example the plethora of different colored fluorescent proteins,
00:04:55.19 that developed by Roger Tsien and co-workers
00:05:00.01 and other labs, and also different color fluorescent dyes.
00:05:04.27 And also we have very chemically specific labeling strategy
00:05:10.06 that allows us to link these probes to biological molecules with high specificity.
00:05:17.01 When you put all three elements together, what you have is the ability to
00:05:22.10 investigate and visualize interactions between different molecules and different gene products
00:05:29.14 in living organisms or living cells.
00:05:33.13 Now powerful as it is, light microscopy also has some disadvantages.
00:05:41.05 And one of them is this: the resolution of light microscopy is limited traditionally to a few hundred nanometers.
00:05:51.26 Well, that might sound small to you, nonetheless, inside the cell we have even smaller objects.
00:06:00.01 For example, the typical size of a small protein is just a few nanometers.
00:06:06.02 So if you want to directly visualize or resolve the interactions and networks
00:06:12.00 between these molecules, ideally we would like to have much smaller resolution, or much finer resolution.
00:06:20.13 And moreover, inside the molecule or molecular complexes are these molecular machines.
00:06:29.10 You even have sub-nanometer type of movement.
00:06:34.25 So because of that we would like to re... or keeping the original advantages of light microscopy,
00:06:44.23 its chemical specificity, its live cell compatibility,
00:06:48.05 but extend the resolution of light microscopy all the way down to the molecular scale.
00:06:55.13 So before I tell you methods that allow us to do that,
00:07:00.01 let me tell you what limits the resolution of traditional light microscopy.
00:07:06.01 And we typically call it the diffraction limited resolution, and here is a little bit of physics.
00:07:12.10 Light is a wave, just like this water wave that I showed you at the shore, somewhere,
00:07:20.15 which I actually don't know where this picture is taken from.
00:07:22.25 But you can see that because it's a wave, it diffracts and gives you these interfering wavy patterns.
00:07:30.27 So when you translate this type of physical phenomena into a microscopic or microscopy terms,
00:07:39.15 what that means is that if we try to focus light with a lens to a very, very small spot,
00:07:47.29 Despite how good of a lens you find, the focal spot nonetheless will have a finite size.
00:07:56.14 And that size is about a couple hundred nanometers in the transverse direction,
00:08:04.10 and even approaching a micron in the longitudinal direction or axial direction,
00:08:10.27 or the direction along the axis where the light travels.
00:08:17.06 So if we imagine using such a big spot to scan the sample,
00:08:24.03 and by the way, if I put it into perspective, and if you think about a protein size over this large scale,
00:08:34.10 the protein would be just a tiny, little bit spots, almost invisible in this spot.
00:08:40.08 So if we use this kind of a focal spot to scan a sample,
00:08:44.28 you can imagine that if you have two features that are widely separated,
00:08:50.27 despite the fact that each of them gives a blurry image,
00:08:54.16 of a few hundred nanometers size, nonetheless, they can be well-separated or well-resolved.
00:09:01.03 On the other hand, if you move these features closer and closer and closer together,
00:09:08.01 eventually their image will merge into one.
00:09:11.13 And that's when you are no longer able to resolve them.
00:09:15.02 And this resolution limit was first recognized by Ernst Abbe, so that is why it is called the Abbe resolution limit.
00:09:22.13 And because of the spot size, the focal spot size,
00:09:26.28 which is also called the point spread function,
00:09:29.19 which is about two hundred nanometers in xy, and 500 nanometers in z,
00:09:35.05 that translates into a resolution of 200 nanometers in xy and about 500 nanometers in z.
00:09:41.13 Okay, so with this in mind you can actually imagine a couple of different types of ways to break this diffraction limit.
00:09:50.24 And one of them is in this approach, called stimulated emission depletion microscopy,
00:09:59.11 or its close relative, the RESOLFT microscopy,
00:10:02.27 which was developed by Stefan Hell at Max Planck Institute,
00:10:11.01 who is really the pioneer in super-resolution microscopy, or sub diffraction limit microscopy.
00:10:17.27 And what they use is using stimulated depletion as a phenomena to effectively trim the focal spot of a microscope
00:10:32.29 into a much smaller one by using a donut shaped stimulated emission depletion beam
00:10:42.10 to trim the fluorescence at the edge of the fluorescence microscope.
00:10:46.01 So with that you can get a substantially smaller spot, and with that spot you can actually scan through small objects
00:10:55.17 inside the cell and get much finer resolution in this case
00:11:00.09 what you see is the image of a mitochondria or several mitochondria.
00:11:05.13 Now also in this spirit, another method developed by Heintzmann and Gustafsson,
00:11:15.09 this is the saturated structural illumination microscopy, so this is better understood if we think in the frequency domain.
00:11:24.08 If we actually use a spatially patterned or periodic pattern of light to interfere or to modulate the sample,
00:11:35.21 and then what you can do is you can beat some of the high frequency information of the sample
00:11:40.11 into a lower frequency information that is detectable by the microscope,
00:11:45.15 and through that, enhance the resolution of the light microscope.
00:11:52.19 And if you translate that into a spatial term, that also effectively shrank the point spread function,
00:12:01.22 or the spot size of the microscope, or the focal spot size of the microscope.
00:12:06.10 But I should say that this is not using a focused spot, this is a wide field imaging technology.
00:12:13.05 So these are the techniques that effectively shrink the point spread function of the light microscope
00:12:21.17 and then use that to break the diffraction limit and achieve a higher imaging resolution.
00:12:27.01 There's another approach that also can allow us to break the diffraction limit
00:12:37.14 and that takes advantage of a single molecule imaging.
00:12:41.20 And we know that even though a single molecule image is diffraction limited,
00:12:49.03 means that the width of this whole image is a few hundred nanometers wide.
00:12:56.04 Nonetheless, in the single molecule field we know that we can pinpoint the center of this image to a much finer resolution.
00:13:09.02 And this localization precision, it depends on the number of photons we detect in this image.
00:13:16.19 If the image width is a few hundred nanometers, and if the number of photons that we detect,
00:13:22.17 for example this N here, it's like 10,000.
00:13:25.15 And then the square root of this N will be about 100.
00:13:30.27 So we can actually pinpoint the center of this image with two nanometer precision, instead of 200 nanometer precision.
00:13:40.27 And this powerful, high precision localization approach can be pretty well understood by this sort of everyday knowledge.
00:13:52.27 You know if we have a very crisp picture of something, for example in this case what we see is Mt. Everest.
00:13:59.19 And we know that the mountain is miles wide, but nonetheless, because of the incredible clarity of the picture
00:14:09.20 or the signal to noise ratio we can pinpoint its peak with just a few meters, not a few nanometers,
00:14:19.19 a few meters or within a meter kind of precision.
00:14:23.12 Now with this very powerful single molecule localization approach
00:14:29.26 scientists in many labs have actually investigated a variety of biological systems.
00:14:36.06 And here what I show you is a beautiful example demonstrated in Paul Selvin's lab
00:14:43.01 where they used this approach to directly visualize
00:14:48.22 the stepping of a motor protein called Myosin V along its actin track,
00:14:55.16 and then they directly visualized this kind of hand over hand,
00:14:59.16 of course in my illustration I was showing the leg over leg type of movement model.
00:15:06.21 Now as powerful as it is, this single molecule localization method does not directly translate into a high resolution image
00:15:19.28 or sub-diffraction limit resolution image.
00:15:22.27 And the reason is that it requires well-isolated images of single molecules.
00:15:29.13 Now if two molecules already give you well isolated images,
00:15:34.08 that means that they can already be resolved by conventional light microscopy.
00:15:38.28 What you typically see in a biological sample is something like this.
00:15:45.14 It is so densely labeled with fluorescent probes and the images of these fluorescent probes
00:15:51.24 severely overlap with each other to a point that you can no longer fit the image of the individual probes
00:16:00.22 and find where their center is.
00:16:02.14 So for example, here can we really imagine fitting this with millions of Gaussians
00:16:09.25 and that is how we actually fit the single molecule image to find the center.
00:16:14.18 Can we really fit this with millions of Gaussian functions and find the center?
00:16:20.09 That would create a lot of propagating error.
00:16:23.02 So how do you overcome this problem?
00:16:27.08 So we and others have come up with this new strategy to overcome this problem
00:16:36.18 and allow us to do sub-diffraction limit imaging was by taking advantage
00:16:44.17 of the photo-switching property of fluorescent probes.
00:16:49.07 So what you see here are probes that can be turned on and off by light.
00:16:56.10 These probes can be switched from a fluorescent state and a dark state.
00:17:00.20 So what we do is we try to go back to the condition where the images of individual molecules
00:17:07.19 are well isolated from each other, so only a sparse subset of them are turned on.
00:17:12.18 And how do we do that? We use relatively weak activation light to activate these probes into...
00:17:18.13 or subsets of these probes... into the fluorescent state.
00:17:24.26 So small a subset now that you can see the green images
00:17:28.28 are again well isolated from each other, and that allows us to pinpoint their center position,
00:17:36.19 and, of course, that is a very big under sampling of the sample.
00:17:42.03 So you overcome that by turning this subset off
00:17:45.16 and then you turn on a different subset and you localize their position, and you iterate this procedure
00:17:54.09 for many, many times until you determine a sufficient number of positions of molecules
00:18:01.27 that allow you to draw the high resolution images
00:18:09.07 or paint the high resolution images of the underlying structure with these molecular positions.
00:18:15.22 Now because this activation procedure is stochastic, every time we activate a stochastic subset of the fluorophores,
00:18:27.19 so we termed this massive stochastic optical reconstruction microscopy, and the short name of it is STORM.
00:18:38.13 And there are, independent to our development,
00:18:44.03 there are two different groups in the world that have come up with a similar imaging concepts.
00:18:50.17 And these are Eric Betzig, Harold Hess, and Jennifer Lipincott-Schwartz,
00:18:57.28 and they call their technology photo-activated localization microscopy, or the name PALM.
00:19:07.01 And another team that is led by Samuel Hess,
00:19:13.22 and he termed his technology fluorescence photo activation localization microscopy. Hence the name FPALM.
00:19:23.22 And these three developments have been published nearly simultaneously.
00:19:32.10 So here is a real example of STORM. You know instead of this structure that is illuminated all the time,
00:19:44.15 what we do here is use low activation intensity and activate only a small subset of a probe at a time,
00:19:54.19 and determine their center position, and plot out the center position.
00:20:00.08 And even in this very short sequence you can already see that the comparison between the conventional
00:20:06.18 image and the STORM image shows a boost of resolution substantially.
00:20:13.20 So in our original paper of STORM we thought that these points of molecules are just painted in the picture.
00:20:26.12 We really wanted to know that they are put in the right position. In other words, we want to validate this approach.
00:20:32.09 So we validated with some biological samples of known shape.
00:20:38.01 And these are the DNA which we tagged photo switchable probes with a known distance,
00:20:45.00 for example in this case 45 nanometers away from each other.
00:20:48.28 And then you can see that because 45 nanometers is so much smaller than the diffraction limit in the conventional image
00:20:57.07 what you see is a blurry spot.
00:20:59.18 Then in the STORM image you can see a sharp contrast. You can see two spots well separated from each other.
00:21:05.28 And similarly in this sample when you have four probes, again a blurry spot of everything merged together,
00:21:14.01 in the conventional image, and in the STORM image you can clearly see four spots.
00:21:20.03 We can not only resolve these close by probes, we can also measure the distance between the probes very precisely.
00:21:28.07 And then when we measured many such samples, we come up with the gray histogram shown in this picture.
00:21:35.11 And then as a final validation point,
00:21:37.27 we compared that with the theoretically predicted distribution of the distance
00:21:44.00 as shown here by the blue curve based on the known basepairs
00:21:49.12 that are separating the probes and the persistence length of DNA.
00:21:53.01 And you can see an almost spot on agreement validating the approach quantitatively.
00:21:59.03 And we can also look at even more complex samples. Here what we see
00:22:03.21 is a filament, a RecA filament, which is RecA protein coating a circular DNA,
00:22:11.01 and again because of the small size of the object you see a blob in the conventional image,
00:22:16.15 and you see a very nice sharp ring well resolved in the STORM image.
00:22:21.21 And we can apply this technology to cellular samples.
00:22:26.04 Here what you see is a STORM image of microtubules inside the cell
00:22:31.01 that is immuno-stained with a dye labeled antibodies.
00:22:35.09 And again if we want to compare the conventional image and STORM image
00:22:43.16 then we can see a drastic improvement of resolution.
00:22:48.10 Where things look relatively smeary and overlapped,
00:22:51.07 here what you see is each and every filament well-resolve from the others.
00:22:59.15 And I also want to emphasize that this is a really, really simple to implement technology.
00:23:05.10 So I'll give you this example, I want to say that actually to do STORM imaging,
00:23:13.04 you don't need special equipment, just a simple epi-fluorescence microscope
00:23:18.19 that many labs already have, with a relatively sensitive camera that has single molecule sensitivity
00:23:25.22 can allow you to do that, in an extremely simple imaging mode, as I will demonstrate with this example.
00:23:32.00 So a few years ago we found this particular family of dyes,
00:23:38.00 the red cyanine dyes such as Cy5 or its structural analog, Alexa647
00:23:44.03 which can actually be toggled between a fluorescent state and a dark state,
00:23:52.05 where the dark state is thiol adduct of the fluorescent state structure,
00:23:57.25 and the molecule can be switched off using red light, and switched back on using the UV light.
00:24:08.04 And it is really a, not only a commercially available probe, but also incredible in its brightness.
00:24:15.21 It allows us to detect 6000 photons per cycle, which makes it one of the brightest,
00:24:22.16 if not the brightest commercially available, photo-switchable probes.
00:24:27.11 And since the number of photons is directly translated into resolution, this also gives you an extremely high resolution.
00:24:36.16 And a very nice property of this is that not only can it be turned back on by UV light,
00:24:44.07 even the red light has a small probability of turning the dark state molecule back to the fluorescent state.
00:24:52.03 Such a small probability, what you see is in your field of view, only a sparse subset of the molecules turn on.
00:25:00.29 So here then the red light continuously on does all three things required for STORM imaging.
00:25:07.20 It excites fluorescence from the molecule.
00:25:09.29 It turns it off, and it also has a small probability to turn it on,
00:25:14.07 and then what you see is molecules blinking on and off
00:25:18.10 in your field of view, and their images are separated at any time so that you can pinpoint their position.
00:25:24.24 So that allows you to actually use a continuously on laser light without any modulation,
00:25:34.19 and a simple epi-fluorescence microscope and that allows you to accomplish a STORM image that looks like this.
00:25:41.27 And to drive the message home, when we took this image we used a single red light,
00:25:47.23 no modulation, and a commercially available Alexa647 labeled antibody.
00:25:55.08 And you can see nonetheless the STORM image gives you a much higher resolution than a conventional image.
00:26:02.03 Now I should say that I didn't actually talk about everything that is needed for this. The equipment is indeed very simple,
00:26:12.21 but with these images which are composed of thousands to millions of localizations,
00:26:19.09 you couldn't possibly hope that you are going to be there actually pinpointing the center of every molecule manually.
00:26:27.00 So you need an automated algorithm, and that many investigators told me could be the largest barrier
00:26:34.26 for them to implement this technology in their own labs.
00:26:38.10 So we can provide solutions to this.
00:26:41.07 We make our image analysis freely available to anybody or any lab.
00:26:50.20 You can just email us and we will send you the software.
00:26:54.16 And the other option if you have the money resources is even easier,
00:26:58.27 and that is that we licensed this technology to Nikon,
00:27:04.05 and then they actually sell a STORM machine that you can just turn the key on and use that to image your sample.
00:27:12.03 And actually there are also other microscope companies that actually sell similar products.
00:27:20.29 So in essence I think that any lab can actually practice this type of technology
00:27:27.07 and get extraordinarily high resolution images in their own hands,
00:27:32.27 either by using the scope that you already have or by buying the commercially available super resolution scopes.
00:27:40.27 And we can also do many color imaging, as I said, multicolor imaging is an advantage of light microscopy,
00:27:51.21 and in this earlier example I showed you
00:27:56.03 a dye that you can use red light to excite and to switch off and then UV light or red light itself to switch it on,
00:28:03.28 and then there are actually different colored photo-switchable probes
00:28:07.25 that emit different colors and then you can then use these to actually do multicolor imaging
00:28:14.19 by differentiating their emission wavelengths
00:28:18.16 or there's also a different mode of multicolor imaging enabled by the fact that
00:28:25.13 we need to photo-switch the molecule,
00:28:27.21 and that is we have these probes that can be activated by light of different wavelengths
00:28:35.08 and if we differentiate the activation wavelength, we can also do multi color imaging.
00:28:41.06 And here is an example of a two color STORM image taken that way.
00:28:47.03 What you see here- the green are the microtubules and the red is another biological structure called clathrin coated pits
00:28:57.25 and to again appreciate the resolution improvement I'd like to superimpose
00:29:05.13 or overlay on this image the conventional image of exactly the same area.
00:29:12.05 And that is what it looks like.
00:29:13.01 So you can see the substantial difference in resolution.
00:29:20.02 And what I found most telling is these images of clathrin coated pits.
00:29:26.14 Now these are true nanoscopic subcellular structures and these are structures used for endocytosis,
00:29:35.29 or uptake of receptors and nutrients in the cell, into the cell.
00:29:41.24 So by true nanoscopic, I meant the diameter of the structure is just about 100 nm.
00:29:49.06 So it is smaller than the diffraction limit, as a result when you see this conventional image,
00:29:56.10 you cannot discern any substructure of this thing.
00:29:59.24 In contrast, this is what the STORM image looked like.
00:30:05.27 Now you can see its clearly spherical morphology.
00:30:09.11 And because what we label here is clathrin, which is a protein coat of these endocytic vesicles,
00:30:18.24 so you would also expect the structure to look like a hollow spherical shell.
00:30:24.24 And if you actually squint at this image, you can almost tell that is hollow.
00:30:30.21 Of course you know it is sort of a funny or ironic way when I try to tell you
00:30:38.10 how good our resolution technique is and ask you to squint at the image.
00:30:43.28 And the reason that it is not entirely hollow is because this is a three dimensional object,
00:30:51.28 and therefore what we see as a two dimensional projection. Even things on the top
00:30:57.07 and bottom will actually project to the center and make it not entirely hollow.
00:31:02.05 What it speaks for is the necessity to do sub diffraction limit imaging or super-resolution imaging in all three dimensions.
00:31:11.04 So how do we do that?
00:31:14.04 Simply speaking it is quite straight forward.
00:31:18.13 You know we accomplished two dimensional super resolution imaging by determining the X and Y position
00:31:26.21 of each molecule very precisely.
00:31:29.13 Well you just look at the Z position, right, and then you can get the three dimensional image.
00:31:34.16 Well, it is a little bit more difficult than so easily put
00:31:41.09 and the reason is because we take these images with a camera and a camera image is like a paper thing,
00:31:47.27 it doesn't have a thickness. Where to you pinpoint that Z position?
00:31:53.11 However, if you've played with microscopes then you can actually potentially even discover the trick yourself.
00:32:00.08 And that is if you look at objects that are well in focus they give you this nice and sharp image.
00:32:07.08 And for molecules or objects that are out of focus, their images look bigger and blurrier.
00:32:15.07 Now typically that is an unwanted phenomenon because that blurs out your image.
00:32:20.15 But nonetheless, it means the shape of the image actually encodes the Z information,
00:32:28.03 so we can turn this around into an advantage instead of a disadvantage.
00:32:34.17 The only problem here is that at the focus, it's a minimum, so the shape doesn't change very sensitively with Z.
00:32:45.18 And there's another problem. Below focus and above focus-they look similar,
00:32:51.08 so you have degenerates, you can not tell negative Z from positive Z.
00:32:55.08 So we can actually kill both of these birds with a really simple stone,
00:33:01.00 and that is we insert a cylindrical lens into the detection path
00:33:06.16 and then the cylindrical lens will bend light only in one direction, unlike the spherical lens that bent light in all directions.
00:33:15.22 And what happens, or as a result then the images of these individual molecules
00:33:22.04 now in general are not spherical, they are ellipsoidal.
00:33:25.21 Okay, and the ellipticity depends very sensitively on the Z direction.
00:33:31.27 And moreover, above focus and below focus are now elongated in two perpendicular directions.
00:33:41.18 And that breaks the degeneracy very nicely. And this is called astigmatism.
00:33:46.26 Using this astigmatism imaging, we can actually, and we first demonstrated 3D STORM in 2008.
00:33:55.21 And I should say that this is just one method that could actually do this.
00:34:01.16 There are a variety of other types of methods that could accomplish this three dimensional information,
00:34:08.10 for example, listed here, bi-plane imaging, double-helical point spread function, interferometry,
00:34:14.22 micromirrors, and maybe even other approaches.
00:34:18.24 So there's a number of ways to pinpoint the position of a molecule with very high precision in all three dimensions.
00:34:27.26 Now using this astigmatism imaging, which is really one of the, perhaps the simplest implementation of all this,
00:34:36.23 and we actually converted the 2D STORM image of the clathrin coated pits
00:34:44.03 like this into the 3D STORM image of clathrin coated pits.
00:34:49.21 And what you see here is that the computer is scanning
00:34:54.06 through the positions of molecules that are in different Z range.
00:34:59.20 Now I should say that this is just virtual scanning,
00:35:02.11 the method itself does not require any mechanical scanning
00:35:05.29 because we determine the X, Y, and Z positions simultaneously just by the shape of the image.
00:35:12.08 And if we ant to see this more clearly, we can, for example,
00:35:17.19 just cut across the plane that is near the plasma membrane,
00:35:23.15 and now you can see that indeed these are hollow ring like structures
00:35:28.08 and you really don't need to squint at it anymore.
00:35:30.03 And if you compare the conventional image and the STORM image
00:35:36.12 you can see how much more information you can get at this molecular scale.
00:35:43.10 Now if we zoom in to one such pit,
00:35:46.26 this is the maximum density projection image of one pit in the XY plane.
00:35:53.22 And if we cut across a plane and take a cross-section near the plasma membrane,
00:36:00.05 you see this nice hollow structure.
00:36:02.07 And if we cut across a perpendicular plane, the XZ plane,
00:36:06.15 in the middle of the pit, and you see this nice cup-like structure
00:36:11.13 and of course, it is called a clathrin coated pit, so it's a pit like structure.
00:36:16.06 So it really recapitulates this type of textbook picture very well.
00:36:20.21 And the resolution of these images if you speak in terms of localization precision
00:36:29.02 in terms of standard deviation, it is less than 10 nm in XY.
00:36:34.25 And less than 20 nm in Z. And in terms of resolution, we think the full width, half max (FWHM)
00:36:42.00 gives a better measure of how far apart the molecules need to be in order of them to be resolved,
00:36:49.04 and that gives you about 20 nm in XY- less than 20 nm in XY- and less than 50 nm in Z.
00:36:59.28 And all around this gives you at least one order of magnitude improvement
00:37:06.18 over widely used confocal microscopy technology.
00:37:12.23 And I said that this technique doesn't involve any mechanical scanning.
00:37:19.16 That's just a little bit of a lie because you can see that there is only a certain range
00:37:28.17 that you can pinpoint at the Z position by this kind of technique.
00:37:34.03 If things are way out of focus, your images get too blurred out
00:37:37.24 and then you just don't have enough signal to noise ratio to pinpoint the center very precisely.
00:37:43.26 So if we want to image a thicker sample, we combine this 3D STORM
00:37:50.09 with a core stepping so that we can image, for example,
00:37:55.00 several hundred nanometers without scanning, and then we move the focus
00:38:00.02 to a different plane and then image another several hundred nanometers
00:38:04.06 and then we step like this until we cover the cell.
00:38:08.24 You know the whole sample in this case, a whole cell
00:38:11.27 with these several hundred nanometer slices.
00:38:16.20 And then here is an example we show of whole cell STORM image,
00:38:20.10 3D STORM image this way, by imaging this object which is mitochondria inside the cells.
00:38:28.09 So what you are going to see is the computer again virtually scans
00:38:32.23 through the depths of the cell and showing you the organelle.
00:38:39.01 And then it will flip around and again virtually scan through in this vertical section.
00:38:46.22 And I should say that since mitochondria again is an object that is
00:38:52.29 comparable, at least its width, to the diffraction limit,
00:38:55.23 seeing these hollow structures is again enabled by super resolution imaging.
00:39:02.25 And here what you see is a more clear illustration of that message.
00:39:07.02 And here is the conventional part, image of part of the cell,
00:39:12.09 and here is the 3D STORM, now the Z information is color coded,
00:39:17.04 and here is a cross section where you can resolve the hollow structure of mitochondria.
00:39:24.10 And we can also do live cell 3D STORM. As I said, live cell imaging is again an advantage of light microscopy.
00:39:35.21 So we implemented live cell 3D STORM for example with this model structure, which is clathrin coated pit again,
00:39:46.03 but this time in two color together with the cargo it carries inside the cell called transferrin.
00:39:53.14 So in terms of labeling, we accomplished this with two tricks.
00:40:01.08 One is so we can directly label the protein in terms of transferrin, which is an exogenously added thing to the cell,
00:40:10.00 so we directly label that.
00:40:11.11 Or for the clathrin, which is an intracellular protein,
00:40:16.27 we use this hybrid fusion approach which here this example is the SNAP technology.
00:40:25.02 developed by the Johnson laboratory.
00:40:28.11 So what's done here is the SNAP protein tag is genetically fused to the target protein,
00:40:35.22 which is transferrin, and the SNAP tag can react with a substrate
00:40:40.25 that is linked to a fluorophore that is suitable for STORM imaging.
00:40:46.13 For example the fluorophore again in this case, is this really nice, wonderful, bright probe, Alexa647.
00:40:56.19 And again in this single snapshot of the live cell image,
00:41:01.00 you can see again this 100 nm object is well resolved into a cup structure.
00:41:07.13 And in this two color image where clathrin is tagged with Alexa647,
00:41:14.10 and the transferrin is directly labeled with Alexa568, you can see the cargo inside the pit.
00:41:21.08 And the resolution needs some explanation.
00:41:26.09 Because I've mentioned the resolution in terms of how precisely you can localize each molecule.
00:41:33.20 Now that's not the only thing that matters to whether the final image has a high definition.
00:41:40.21 We also need enough localizations to lay out this structure.
00:41:46.03 You can imagine that with such a nice pit structure,
00:41:48.15 if I only have three dots in it, no matter how precisely each dot is localized
00:41:53.23 you will not paint out or outline the complex structure.
00:41:58.06 So we need a sufficient density of the localizations.
00:42:03.25 And the average distance between nearby molecules if you double that,
00:42:09.09 that gives you a nice resolution measure called the Nyquist resolution.
00:42:14.14 And in these images we accomplished about 20 nm, 24 nm Nyquist resolution.
00:42:22.13 When you combine thiat with the localization precision of individual fluorophores,
00:42:27.27 that gives you about 30 nm in XY and 50 nm in Z.
00:42:35.26 Again, about an order of magnitude better than confocal light microscopy.
00:42:40.13 And what is really good is the time resolution because people had...
00:42:48.14 because of the fact that this imaging approach has to accumulate many, many frames
00:42:54.22 and then sort of super impose the localizations from individual frames.
00:43:00.12 People had expected it to be a slow imaging approach.
00:43:06.05 And indeed it is slower than methods like confocal light microscopy.
00:43:10.07 Nonetheless, here we demonstrated that this kind of super-resolution image
00:43:15.03 can be acquired in something of a second.
00:43:19.23 And sometimes we even accomplish sub second time resolution.
00:43:22.26 And that really allows us to apply this technology to study dynamics
00:43:30.08 inside the cell of many different processes.
00:43:36.05 And another important thing to mention is
00:43:40.07 although it takes about a second or so, or sometimes longer, sometimes shorter,
00:43:46.07 to acquire or accumulate this whole high resolution image,
00:43:51.27 we can also combine that with single particle tracking and track individual probes
00:43:59.07 or molecules inside the structure with millisecond time resolution.
00:44:04.08 So when you combine this, it really offers a very powerful, fast approach
00:44:11.27 for studying dynamics of structures inside the cell.
00:44:17.27 I should say that this very fast imaging is really enabled by
00:44:26.16 using these probes like the photo-switchable dye Alexa647
00:44:32.21 which is very bright and which can be switched on and off very rapidly.
00:44:38.18 And with other probes that we tried, for example photo-activatable fluorescent proteins,
00:44:45.27 we have a resolution, a time resolution, in the range of a few 10s of nanometers,
00:44:51.29 which is also not that bad, and can be used to study a lot of dynamic processes inside the cell.
00:45:01.26 So as I mentioned, one of the key elements to STORM is the photo-switching capability of these probes.
00:45:12.29 You have to be able to turn on just a subset of the molecules at will.
00:45:18.13 And people have wondered whether that limits that application range of STORM microscopy.
00:45:27.06 And my answer to that question is not really,
00:45:29.15 because actually as we had already stated in our very original STORM paper
00:45:36.26 a number of years ago that although in this first work
00:45:41.27 we demonstrated STORM with these photo-switchable cyanine dyes,
00:45:46.16 the method is nonetheless applicable in principle to any suitable
00:45:53.16 optically switched fluorophores could be used.
00:45:56.29 And it is applicable to other photo-switchable fluorophores such as fluorescent dyes and fluorescent proteins.
00:46:04.29 And indeed in the past few years, large numbers of photo-switchable probes
00:46:15.11 including both fluorescent proteins and fluorescent dyes
00:46:20.14 have demonstrated to be suitable for the STORM and PALM and fPALM type of approach,
00:46:29.29 both by my laboratory and a number of other laboratories.
00:46:35.29 And even some of the conventional fluorescent proteins and the conventional fluorescent dyes
00:46:47.08 have been shown to be able to be switched between a fluorescent state and a dark state under illumination.
00:46:56.12 A phenomenon termed blinking.
00:46:59.17 In other words these conventional probes also can be photo-switchable,
00:47:04.01 and they have also been demonstrated to be useful for STORM imaging.
00:47:11.05 So this means that a wide variety of biological structures and biological systems
00:47:19.01 can be investigated by the STORM type of super resolution imaging
00:47:26.27 as we have demonstrated here a wide variety of cellular structures has been demonstrated,
00:47:36.12 or has been imaged to sub diffraction limit... sub diffraction resolution using STORM.
00:47:46.09 So in part 2 I am going to actually give you two such examples of applications of STORM
00:47:54.25 that has generated interesting and new biological insights.
00:47:59.11 But before I end this part, let me thank the people
00:48:05.04 who participated in the development of the STORM technology in my lab.
00:48:12.08 And the original development is largely due to these three individuals highlighted
00:48:19.20 in this section of the former group members.
00:48:24.19 And these are Dr. Bo Huang, who is now an Assistant Professor at UCSF,
00:48:34.09 and Dr. Mike Rust, who is a Assistant Professor now at University of Chicago.
00:48:44.14 And last, but definitely not the least is a former graduate student, Mark Bates,
00:48:51.13 who is now a postdoc at Max Planck Institute with Stefan Hell.
00:48:57.02 And these early developments are with the help of two group members
00:49:03.00 that are still currently in the group, and that is Wenqin Wang and Graham Dempsey.
00:49:06.29 And the live cell work is largely due to these two individuals,
00:49:13.00 and that is Sara Jones, a very talented graduate student,
00:49:18.19 and a talented postdoc, Sang-Hee Shim.
00:49:21.29 So with this I will end part 1, and I will invite you to part 2
00:49:32.05 where I talk about the example applications of STORM and part 3
00:49:38.11 where I will talk about single molecule FRET and applications of single molecule FRET in my laboratory.

Part 2: Applications of Super-resolution STORM

00:00:00.14 Welcome back to my iBioSeminar series part two.
00:00:05.26 Again, I am Xiaowei Zhuang. I am a professor in the department of Chemistry and Chemical Biology
00:00:13.06 and the Department of Physics at Harvard University.
00:00:15.23 And an Investigator of the Howard Hughes Medical Institute.
00:00:20.25 Now in part one, I introduced super resolution fluorescence microscopy.
00:00:30.08 And in particular the single molecule based super resolution imaging method, STORM.
00:00:37.07 And in this part I will give you a couple of example applications of STORM in microbiology and neurobiology.
00:00:51.13 So I will up front say that in this application
00:00:57.10 we use photoswitchable fluorescent proteins to do the STORM imaging.
00:01:04.02 And in this application we use photoswitchable dyes to perform STORM imaging.
00:01:13.19 So first our application in microbiology.
00:01:17.26 And one area of application that I am very excited about is bacterial imaging.
00:01:26.01 Now bacteria, these are tiny cells, and if you look at their width,
00:01:31.26 the width is just several hundred nanometers, or sometimes approaching a micron or so,
00:01:38.09 and that is nothing but just a few times the diffraction limited resolution size.
00:01:45.26 So because these cells are so small, it has actually been traditionally difficult to image them.
00:01:54.26 And because of this difficulty of imaging, our understanding of the cell biology of bacteria actually lags far behind
00:02:06.21 our understanding of the cell biology of eukaryotic cells, for example mammalian cells,
00:02:13.12 which is quite funny and ironic considering that these are arguably simpler cells.
00:02:20.27 So super resolution fluorescence microscopy offers an ideal approach to solve this bacterial imaging problem.
00:02:31.13 So here I am just going to give you one example where we study the nucleoid
00:02:37.12 and chromosome organization of E. coli
00:02:42.01 using STORM in collaboration with my colleague Sunny Xie at Harvard.
00:02:50.06 So we know that bacteria is small, and its DNA structure inside the bacteria is even smaller.
00:03:02.28 In fact if you spread out the DNA of an E. coli cell completely,
00:03:09.01 the contour length is almost 1 millimeter, but it has to be compact
00:03:14.14 into a much smaller structure, a submicron structure, called the nucleoid.
00:03:19.10 And this compaction is facilitated by a number of different proteins in the bacteria
00:03:27.09 among which are these major nucleoid associated proteins in E. coli.
00:03:32.13 I listed 5 of them here, HU, H-NS, IHF, Fis, and StpA.
00:03:41.00 And we image their distribution using STORM, and here for example
00:03:46.19 what you see is actually a live cell STORM image of HU in an E. coli cell.
00:03:54.09 And you can see that this HU molecules pretty much are scattered throughout the entire nucleoid.
00:04:02.06 And here shown on the left is the bright field image of the E. coli cells.
00:04:08.21 And this phenomenon is also found to be true
00:04:12.21 for these other NAP or nucleoid associated proteins like Fis, IHF, and StpA.
00:04:24.03 In contrast, HNS distribution looks drastically different.
00:04:31.29 Instead of being spread out throughout the nucleoid,
00:04:35.12 they form these tight clusters inside the cell.
00:04:39.11 And as you can see, that these clusters are much better resolved in the super resolution STORM image
00:04:46.03 as compared to the conventional image.
00:04:49.19 These clusters are quite static, or they don't move around that much.
00:04:55.02 You know here what you see is actually a live cell movie of HNS.
00:05:00.12 It almost looks like a still image because they are not really moving around that much.
00:05:04.17 And by quantifying these types of clustered structures in many cells,
00:05:12.02 we found that these HNS clusters are organized in a very regular way in two different senses.
00:05:21.26 One is you can see the number of clusters inside the cell.
00:05:27.08 That actually increases in general with the length of the cell.
00:05:33.00 The longer the cell, you see more clusters.
00:05:36.27 And for the newly dividing cell, under conditions that we know
00:05:42.01 that there is only one copy of chromosome inside the cell
00:05:46.21 we found basically two such clusters for example here in this image
00:05:52.08 and then you can see here in this bar graph.
00:05:54.26 And that is one way of regulated organization.
00:05:59.26 And the other one is their position.
00:06:02.15 You know, I won't belabor you with these distribution histograms,
00:06:08.09 but just to tell you that by quantitative characterization of many such cells
00:06:13.12 that we found that these two clusters tend to be positioned
00:06:17.19 along the long axis of the cell at the quarter cell positions.
00:06:23.07 And how do they, what is the mechanism that they cluster?
00:06:30.03 To investigate that we did mutation studies so HNS molecules is known to have this type of structure,
00:06:41.29 where it has two domains, a DNA binding domain near the C-terminus,
00:06:46.27 and a dimerization domain at the N-terminus.
00:06:51.07 So to see how they effect the clustering, what we did is we used
00:06:56.18 this previously established mutation L30P to actually disrupt the dimerization tendency of the molecule.
00:07:08.05 And once we introduced this mutation into endogenously expressed proteins
00:07:15.23 by chromosome insertion, you can see that the clusters are no longer there
00:07:21.04 and the molecules are now again scattered throughout the nucleoid, just like HU.
00:07:26.20 And then when we use this previously established mutation that actually disrupts the DNA binding
00:07:35.29 we found much fewer localizations,
00:07:39.00 telling us that these localizations we see are indeed due to DNA bound molecules.
00:07:46.24 So this tells us that the cluster formation is due to the interaction,
00:07:54.00 the self interaction, between the N-terminal domain of the molecules that are bound to DNA.
00:08:01.22 Now functionally HNS is a global transcriptional silencer.
00:08:08.25 It represses about 5% of all E. coli genes, a total of around 200 of them.
00:08:15.14 So to see whether the cluster, how are they related to the repressed genes,
00:08:23.08 we actually looked at two color imaging of HNS clusters and the gene locus.
00:08:31.04 And specifically three of them, the hdeA gene locus, and the hchA gene locus
00:08:38.03 and these are genes that are known to be... their expression is repressed by HNS.
00:08:44.08 And as we also show here with RT-PCR that indeed their expression level is substantially reduced in wildtype cells
00:08:55.00 as compared to HNS deleted cell. And as a control we looked at the lacZ gene locus.
00:09:03.28 And lacZ, you can see its expression level is not repressed by HNS.
00:09:09.21 And then here is the two color super resolution images of HNS in purple
00:09:17.16 and the gene locus in green.
00:09:20.15 And as you can see on these left two panels there is substantial co-localization between the gene locus
00:09:28.11 and the HNS cluster while there is no such substantial co-localization for this control gene.
00:09:35.23 Actually by sequestering this gene locus into the cluster we also found
00:09:42.28 that the gene locus has been moved by substantial distances inside the nucleoid by these clusters
00:09:52.11 as demonstrated in these distribution maps where we compared the distribution of the gene locus
00:10:00.03 in wildtype cells and in cells where HNS is deleted.
00:10:06.20 And here when you see two lobes and four lobes that doesn't mean there are four such gene locus inside the cell.
00:10:14.28 They just mean these are the positions where the gene locus is most likely to be.
00:10:20.17 And you can see that the most likely position has been moved substantially by the HNS clustering effect
00:10:30.13 in both of these two cases, while in the control genes
00:10:34.23 you can see that the most likely positions are not moved too much when you delete HNS.
00:10:43.10 And as I said since HNS is a global transcriptional silencer which represses about 200 genes, and they actually
00:10:53.10 have even more than 200 binding sites that are distributed throughout the E. coli genome broadly.
00:11:01.27 So by sequestering these sites into the clusters
00:11:06.08 we think that HNS has a global organization and compaction
00:11:12.07 or condensation effect on the bacteria or E. coli chromosome.
00:11:17.22 To see that these genes are indeed brought into proximity in a more high throughput way
00:11:24.09 we utilized this technology or this method called chromosome conformation capture
00:11:30.18 developed by Nancy Kleckner and Joe Dekker a number of years ago at Harvard.
00:11:38.22 And in this technology the gene locus, or the DNA locus, that are brought into proximity by proteins
00:11:49.22 are probed by crosslinking and then later on examined by PCR for example.
00:11:57.24 So we used this to see if among these locus that are regulated or the genes that are regulated
00:12:06.23 by HNS whether they are indeed brought into proximity.
00:12:11.11 So here in this plot what you see the black circles are the genes that are regulated by HNS
00:12:19.03 and as a comparison we also are putting these control loci by the red crosses.
00:12:26.06 And then we probed the pairwise proximity by their crosslinking frequency, their crosslinking efficiency.
00:12:35.09 And as you can see here on this bar graph among the control pairs
00:12:41.18 between the negative control size or between the control size
00:12:46.22 and the HNS regulated site, these control pairs give you very low crosslinking frequency,
00:12:54.12 while in contrast the pairs that are formed between two HNS regulated loci
00:13:01.10 give you much higher crosslinking efficiency.
00:13:05.02 And moreover, this crosslinking frequency is dependent on HNS
00:13:10.25 because if we delete HNS in the cell, you can see the crosslinking frequency drop down to the control level.
00:13:18.27 So with this we have this model picture of how HNS
00:13:28.14 proteins could help organize chromosome inside the cell.
00:13:33.23 And that is they form tight clusters in the nucleoid,
00:13:36.27 and quantitatively from each copy of chromosomes there are about two such clusters.
00:13:44.27 And they form by oligomerization of the DNA bound HNSs through their N-terminal domain interaction
00:13:53.17 and by sequestering many, many gene locus or DNA locus, broadly distributed on the chromosome
00:14:01.21 into these tight clusters we think that this molecule
00:14:07.02 helps DNA folding and condensation by serving as organization centers
00:14:14.12 that anchor and juxtapose numerous of DNA... HNS binding sites on the bacterial chromosome.
00:14:25.11 So that is the first example and the second example as I mentioned earlier is in neurobiology
00:14:37.01 and in particular in this example we are investigating the subsynapse structures
00:14:45.10 or the molecular architecture of synapses.
00:14:48.11 together with my colleague Catherine Dulac at Harvard University.
00:14:55.08 As we know that synapse structures are the fundamental functional unit
00:15:01.22 inside the brain that connects neurons and their target cells.
00:15:09.00 And I like to say that synapse is also an ideal subject to investigate with super resolution fluorescence microscopy.
00:15:18.10 And the reason is it is also tiny.
00:15:21.29 The size of a synapse is several hundred nanometers, again not much larger than the diffraction
00:15:31.07 limited resolution of conventional light microscopy.
00:15:34.17 Inside this tiny structure resides hundreds of different types of proteins.
00:15:42.01 Some are existing in high copy number and therefore you are seeing
00:15:48.08 actually thousands or more molecules densely packed
00:15:55.07 in this basic functional unit in the brain that all these molecules are working in a concerted manner
00:16:02.27 to mediate the neurotransmitter mediated signal transmission between neurons and their target cells.
00:16:11.04 So it is thus essential that we understand how these molecules are organized and how they interact with each other
00:16:20.12 in order to understand the functional mechanism of this structure.
00:16:25.12 And this is very difficult to investigate with the conventional light microscopy tool as I illustrate in this image,
00:16:38.18 this conventional image of a synapse structures which are labeled with two proteins.
00:16:45.13 One is the presynaptic scaffolding protein Bassoon in red and one is the postsynaptic scaffolding protein Homer1.
00:16:54.00 And as you can see in this image, there is severe overlap between these two
00:16:59.15 structures and also in places multiple synapses are overlapping with each other.
00:17:06.12 And here is the STORM image of the same field of view,
00:17:12.11 and let me move this way so you can see the STORM image more clearly.
00:17:20.02 And then if I zoom in to a smaller region you can even more clearly appreciate
00:17:25.19 the difference between the conventional and STORM image
00:17:28.24 where these two synapses are mistaken as one in the conventional image.
00:17:34.06 Here we resolve it very clearly with the STORM image.
00:17:39.27 And then also you can see the clear separation between the pre and post synaptic scaffolding proteins.
00:17:46.11 Now because this is a three dimensional imaging method,
00:17:51.01 you can actually see the separation of these two structures even more clearly if I show you this movie
00:18:00.26 that rotates and then shows you all 3D perspective of the substructure
00:18:09.04 and then you can see that at the cross section view
00:18:12.19 you can see the clear separation of the presynaptic scaffolding protein
00:18:16.26 and the post synaptic scaffolding protein separated by the synaptic cleft.
00:18:21.05 And here is the en face view showing you how they are distributed laterally.
00:18:25.20 And we can do three color STORM imaging, and here adding the third color actually had substantial capability for us
00:18:40.11 to determine the molecular architecture of the synapse.
00:18:43.25 So here is our experimental design.
00:18:46.16 We used two such markers, again Bassoon and Homer1, to set the reference frame.
00:18:53.20 And then we systematically swapped out the third marker.
00:18:59.03 So for example in this image what you see is GluR1 which is a subunit of the AMPA receptor
00:19:05.17 and as expected it is distributed in the synaptic cleft.
00:19:12.06 And then if we keep changing this third marker then we can actually determine
00:19:16.23 the position of many different synaptic proteins in this framework set up by Bassoon and Homer1.
00:19:25.17 And then here is, using this approach we determined the position
00:19:32.03 of ten or so synaptic proteins along the trans-synaptic axis like this direction.
00:19:42.19 And here you can see that such an architectural map really gives you very rich information
00:19:49.21 about how proteins are distributed and laid out in the synapses.
00:19:56.23 And I won't go into explaining the implication of each individual one due to time sake.
00:20:04.29 And we can not only determine the axial position of these molecules along the trans-synaptic axis
00:20:12.11 we can also because this is a 3D imaging approach to have an en face view
00:20:20.05 to let us determine this lateral distribution of molecules along the synaptic cleft.
00:20:25.19 And for example we looked into the lateral distribution of neurotransmitter receptors,
00:20:34.28 which are of particular interest because they could effect the transmission property of the synapses.
00:20:42.04 And here what you see is the distribution of GluR1, which is an AMPA receptor subunit
00:20:51.08 and NR2B which is an NMDA receptor subunit with respect to the postsynaptic density marker Homer1.
00:21:00.20 And we can see both central synaptic distribution where the receptors are almost exactly overlapping
00:21:08.16 with the postsynaptic density and a perisynpatic distribution
00:21:14.01 where the receptor forms an annular ring around the postsynaptic density.
00:21:19.20 Now we can not only get this kind of lateral distribution,
00:21:23.18 we can also get quantitative information on the composition of receptors
00:21:29.13 in the synapses which is another important property that could determine the transmission property of synapses.
00:21:37.07 Now if you look at these two images you might even think of these as
00:21:44.04 two color images because you can see green and blue,
00:21:47.11 but in fact these are actually three color images
00:21:50.19 where all three probes, Homer1, GluR1, and NR2B are simultaneously probed.
00:21:56.29 There is only largely two colors because this protein NR2B is largely missing in these synapses.
00:22:08.16 In the same brain area we can also see distributions or composition like this
00:22:14.16 where you can see a large amount of red signal indicating
00:22:19.24 actually a lot of NR2B subunits are located in these synapses.
00:22:27.15 So it tells you that the composition of synapses can be highly heterogeneous, even in the same brain area.
00:22:34.27 By quantifying the neural transmitter receptor composition over many synapses in a couple of brain regions
00:22:44.13 we actually come up with a pretty surprising discovery.
00:22:48.25 And that is that we found that in this adult accessory olfactory bulb which is short named here as AOB,
00:23:00.24 we found a large population of synapses that predominantly has NR2B,
00:23:11.19 but has very few GluR1. And if we further look at the
00:23:18.22 postsynaptic density marker, Homer1, and other postsynaptic density proteins,
00:23:25.00 we found that these synapses that predominantly has this NMDA receptor subunit NR2B
00:23:33.07 but lack largely the AMPA receptor subunit GluR1,
00:23:38.24 they also have a very small amount of the postsynaptic density protein.
00:23:45.12 So all of these put together, these three signatures, low AMPA receptor content,
00:23:51.15 high NMDA receptor content, and low postsynaptic density content
00:23:57.17 could mean that these are likely immature synapses, or silent synapses.
00:24:05.07 And I say this is a surprising discovery because this is found in
00:24:12.10 adult mice in their accessory olfactory bulb or AOB regions.
00:24:17.23 And while these immature or silent synapses are known to be
00:24:22.29 present in large number in the early developmental stage
00:24:28.23 and they are known to over time mature and disappear, get converted into a mature synapses
00:24:38.05 before the animal reaches adulthood.
00:24:41.12 And indeed in the MOB region, the main olfactory bulb region, you see very few of such synapses.
00:24:48.15 So to see whether these potentially immature or silent synapses could be converted into mature ones
00:24:58.11 by stimulation, we utilized this wonderful technology, optogenetics technology,
00:25:07.10 pioneered by Karl Deisseroth's lab at Stanford University.
00:25:13.14 So here we took advantage of a mouse line that is expressing channel rhodopsin
00:25:22.06 in the vomeronasal organ sensory neurons which synapse with the mitral cells in the accessory olfactory bulb.
00:25:30.19 And this mouse line is generated by my colleague Venky Murthy's lab.
00:25:39.05 And then we used light to stimulate these AOB synapses
00:25:44.25 and we wanted to see after stimulation whether these synapses could mature
00:25:49.28 into a more mature synapse.
00:25:54.02 And then here what you see is the blue distribution is before the light stimulation,
00:25:59.22 and the green distribution is after stimulation.
00:26:03.05 You can see indeed, these synapses that have low GluR1 content
00:26:09.06 indeed matured towards this side by accumulating GluR1 and also by decreasing their NR2B content.
00:26:18.06 So with this example I hope to have shown to you that 3D STORM allows quantitative characterization
00:26:29.07 of molecular architecture of cellular structures as small as synapses.
00:26:36.11 And of course, it can also be applied to many other tiny cellular structures
00:26:40.14 that are difficult to study by conventional approaches.
00:26:44.12 And indeed we have already made discoveries with this kind of super resolution approaches.
00:26:53.14 So with that, let me again acknowledge the people who have done the work.
00:27:01.13 And the bacterial imaging work is done by a graduate student in the lab Wenqin Wang,
00:27:11.08 as I said, in collaboration with my colleague Sunny Xie's lab,
00:27:16.01 and his graduate students Gene-Wei Li and Chongyi Chen.
00:27:21.02 And the synapse work is done by Bo Huang when he was in the lab
00:27:29.07 in collaboration with my colleague Catherine Dulac's laboratory and her former postdoc Adish Dani.
00:27:38.13 And this concludes the second part of my iBioSeminar series
00:27:46.04 and then the third part will be on single molecule FRET studies
00:27:53.14 of a couple of biological systems that are involved in gene replication
00:27:59.15 and expression and regulation and so on. Thank you.

Videos in this Talk

Part 1: Super-Resolution Fluorescence Microscopy
Audience:
- Student
- Researcher
- Educators of Adv. Undergrad / Grad
Part 2: Applications of Super-resolution STORM
Audience:
- Student
- Researcher
- Educators of Adv. Undergrad / Grad

Speaker: Xiaowei Zhuang

Total Duration: 1:18:17

Recorded: July 2011

All Talks in Biophysics

Talk Overview

Zhuang begins her lecture by explaining that the resolution of traditional light microscopy is about 200 nm due to the diffraction of light. This diffraction limit has long hampered the ability of scientists to visualize individual proteins and sub-cellular structures. Xiaowei Zhuang discusses how sub-diffraction limit, or super-resolution, microscopy techniques, such as STORM, allow scientists to obtain images with greater resolution and to study cell dynamics in greater detail.

Speaker Bio

Xiaowei Zhuang

Xiaowei Zhuang is Professor of Chemistry and Chemical Biology, and Professor of Physics at Harvard University and a Howard Hughes Medical Institute Investigator. She received her B.S. in physics from University of Science and Technology of China and her Ph.D. in physics from the University of California, Berkeley. As a post-doctoral fellow with Steven Chu… Continue Reading